A New Mathematical Modeling Technique

for Pull Production Control Systems

O. Srikanth1, B. V. Ramana Murty2 & A.V. Sita Rama Raju3
1

Mechanical Engineering Department, Dhanekula Institute of Engineering and Technology,

Vijayawada, A.P, India
2
Mechanical Engineering Department, Gayatri Vidya Parishad College of Engineering,
Visakhapatnam A.P, India
3
Mechanical Engineering Department, Jawaharlal Nehru Technological University, Hyderabad, A.P, India
E-mail : 1muralisrikanth11@gmail.com , 2srisai666@yahoo.co.uk, 3avsr_raju2005@yahoo.co.in

Abstract The Kanban Control System widely used to

control the release of parts of multistage manufacturing
system operating under a pull production control system.
Most of the work on Kanban Control System deals with
multi-product manufacturing system. In this paper, we
are proposing a regression modeling technique in a multistage manufacturing system is to be coordinates the release
of parts into each stage of the system with the arrival of
customer demands for final products. And also comparing
two variants stages of the Kanban Control System model
and combines with mathematical and Simulink model for
the production coordination of parts in an assembly
manufacturing systems. In both variants, the production of
a new subassembly is authorized only when an assembly
Kanban is available. Assembly kanbans become available
when finished product is consumed. A simulation
environment for the product line system has to generate
with the proposed model and the mathematical model have
to give implementation against the simulation model in the
working platform of MATLAB. Both the simulation and
model outputs have provided an in depth analysis of each
of the resulting control system for offering model of a
product line system.

like controlling the inventory especially controlling

Work in Process (WIP) is the primary factor.
In general, pull systems use local information
about the status of production and inventories to
authorize the order releases, while push systems use
global information including the firm and forecasted
customer orders, lead times, capacities, etc In a pullbased supply chain, procurement, production and
distribution are demand-driven rather than to forecast.
A pull strategy is a make-to-order production
strategy but not produceto-order like a push production
system. Pull production systems are applied to that
portion of the supply chain with the advantages like
where demand uncertainty is high , Production and
distribution are demand driven , almost no inventory ,
response to specific orders , Point of sale (POS) data
comes in handy when shared with supply chain partners,
Decrease in lead times .
An important question when implementing a pull
control policy is how to define its parameters to achieve
a good tradeoff between the customer service levels and
the inventory costs. For this purpose, performance
evaluation of the Kanban control systems for assembly
systems should be performed using simulation or
analytical approximate techniques.

Keywords pull production control, Kanban systems,

Mathematical Model.

I.

INTRODUCTION

In recent years, manufacturing firms are facing

great challenges to minimize the total production costs
of various products because of their shorter product life
cycle times and varying customer interests and demands
especially under the global competition. Therefore to
retain their competitive advantages, it essential for the
manufacturing firms to adapt to the changing
environment by promoting continuous improvement.
This improvement includes various number of factors

Fig. 1: Structured Flow of Kanban Card

ISSN : 2319 3182, Volume-2, Issue-4, 2013

International Journal on Theoretical and Applied Research in Mechanical Engineering (IJTARME)

Spearman, M.L et al. have proposed single-class kanban

controlled manufacturing system. Order release
mechanism as in classical MRP was sensitive to the
changes in the lead time according to the concept
of lead time syndrome. The problem was modeled as
a two-dimensional Markov chain, and it was solved
explicitly by using matrix geometric techniques.
Comparative analysis was done between static and
dynamic lead time quoting procedures, and
significant cost benefits of the dynamic procedure
were shown under various scenarios. Guidelines for
setting design parameters such as the number of
kanbans and the frequency of updating the lead
time were provided through numerical tests. Arianna
Alfieri et al.
have proposed an approximate
representation for a class of production systems
characterized by several stages, limited buffer capacities
and stochastic production times. The main advantage of
the proposed formulation was that it preserves its
linearity even when used for optimization and, for such
a reason, it could be adopted in simulation optimization
problems to reduce the initial solution space. The
approximate formulation was applied to relevant
problems such as buffer capacity allocation in
manufacturing systems and control parameters setting in
pull systems.

control problem. The existing NDP algorithms and

SBM-PIM are numerically compared with a traditional
UMDP algorithm for a single-stage JIT production system. It was shown that all NDP algorithms except the
SBMPIM fail to converge to an optimal control.
From the review, it can be seen that most of the
earlier research works have performed the Pull-control
kanban system improvement process using regression
modeling technique or adding some pre-processing
method to the kanban card model. The Kanban model
has been developed to preserve the linearity of the
system, but here nonlinearity among the input and
output plays major role. Moreover, the process modeling
has not been enabled with regular optimization process.
Both design and development stages of current
manufacturing process modeling systems suffer because
of the aforesaid key issues. Therefore the development
and operation of the proposed a model firstly for the
same
simulation and mathematical modeling is
presented. The results at the end show the performance
of the proposed mathematical model.

P. Lavoie et al. have proposed a technique of

production lines composed of multiple unreliable
workstations producing a single part type with holding
and shortage costs was considered. Using simulation,
design of experiments and response surface
methodology, various homogenous lines controlled by
Kanban, CONWIP and the hybrid Kanban/CONWIP
control mechanisms were optimized and then compared.
It was observed that the hybrid mechanism always
outperforms CONWIP and Kanban when storage space
and inventory costs were considered explicitly.
However, hybrid was equivalent to CONWIP when
storage space costs were not considered explicitly but
rather aggregated with the inventory costs in the holding
costs as usually found in the literature. Furthermore,
with the increase of storage space cost and the number
of workstations, Kanban outperforms CONWIP.

Fig. 2: Block Diagram of propose modeling a Kanban

Production System

II. THE PROBLEM STATEMENT

Katsuhisa Ohno have discussed a discrete-time

optimal control problem in a multistage JIT-based
production and distribution system with stochastic
demand and capacity, developed to minimize the
expected total cost per unit of time. This problem could
be formulated as an undiscounted Markov decision
process (UMDP); however, the curse of dimensionality
makes it very difficult to find an exact solution. They
proposed
neuro-dynamic
programming
(NDP)
algorithm, the simulation-based modified policy
iteration method (SBMPIM), to solve the optimal

Proposed Technique for Modeling of Kanban Control

System:
In our proposed kanban control system, we use the
number of cards in use is fixed as K. Customer demand
drives the inventory system. Each part is attached with a
kanban. When a customer demand arrives, the finished
part is released to the customer and the kanban attached
to that part is transferred upstream for initiating the
production. In this paper, we have to use regression
ISSN : 2319 3182, Volume-2, Issue-4, 2013

International Journal on Theoretical and Applied Research in Mechanical Engineering (IJTARME)

analysis over the subjected product line system and

hence appropriate mathematical model.

Product Assemble System with Kanban Control:

To release a material part from production process
into the assemble process of a stage, one requires the
following:

The Model of Structured Input Boundary:

In this proposed production system mechanism and
Kanban Control System (KCS) are modeled and
simulated to exploit the combined advantages and also
to study their effect in a typical manufacturing
environment. For the proroposed modules of the
Kanban Control System Simulation studies were
performed using MatLab (SimuLink) to evaluate the
release of components to the downstream. The basic
idea is to impersonate the necessary machinery or
resources according to the database. Therefore, in the
initialization of the simulation model, the model is
parameterized by an interface to the database according
to the requirement.

There must be at least one finished part in the

output buffer in the upstream stage;

There must be atleast one demand to release a ; and

There must be atleast one production authorization

to release a new part into the present stage.

In final assembly section will provide the final

product to the inventory department. Any
nonconformities information will cause a flow effect
through the production system causing upstream work
station to inventory stocks. Since each assembly
schedule must be very similar to all other production
schedules, it is essential that it is possible to freeze the
master production schedule for a fixed time period. A
logical conclusion from this is the need for balancing
between all operations in order to synchronize the
starting and ending of work process. This ensures that
parts are fed to the assembly line at the same rate as they
are consumed.

Raw material Supplier:

The initial number of raw parts in the raw parts
container (RM1 and RM2) the arrival the new parts into
store database. Each container these parts are stored in
to order to production that authorizes the production
workstations. The demand is issued into the system in
the form of order in normal distribution at the last stage
of the each production control system. The maximum
capacity of this demand input queue is set in orders.
This number forms a limit on the system when the
system is operating in an inefficient manner. When a
customer demand arrives to the system, it is
immediately transmitted to next workstation adding one
demand to the container of queues.

Proposed Technique for Modeling of Kanban Control

System with Mathematical Formulations:
After the Simulation model phase, the mathematical
model is developed for Kanban System. We consider a
manufacturing system in which the production of parts
proceeds in stages. Each stage is a production or an
Inventory system made of a manufacturing process and
an output buffer. The manufacturing process consists of
a single machine or a sub network of several machines.
We developed a mathematical model for Kanban
system.

Production System with Kanban Control:

After receiving materials we have to proceed for
next level. Here we customize kanban card in material
container, and send to the final assembly section for the
end product and, with this knowledge, it controls what is
produced in the total manufacturing system using the
following procedure.

In manufacturing system having N cells in

series M n consist of different types of manufacturing

Production with kanbans card:

process with an output buffer. Thus, we have a finite set

of row materials input, Number of WIP Kanban
Assemble indexed is n with each material part assigned
to a machine (RM1 ----RMn). Here, In the KCS each
stage i has K i kanbans a i that are used to authorize

Production kanbans card is define the quantity of

the supplier materials part that the producing work
center should manufacture in order to replace those,
which have been removed with kanban card. And after
the supplying the materials part empty the container will
moved from assemble section to production section.

the production of next

mathematical model is,

Withdrawal kanbans:

Mi

In this section, we have Withdrawal kanban cards

from empty container and that the subsequent process
should withdraw from the preceding work center. Each
card circulates between two work stations only.

stage.

Pi ns
t T

Pi nsi ns ti ns Ti

The

developed

(1)

(2)

ISSN : 2319 3182, Volume-2, Issue-4, 2013

International Journal on Theoretical and Applied Research in Mechanical Engineering (IJTARME)

st sT
i

f li = End timing of the assemble timing

(3)

In this paper, the initial number of raw parts in the

raw parts buffer Ri and the arrival process of new

Pi = It is new material part for the production step.

materials parts into P1 fall outside the scope of the

control mechanism. Each material part M i has

Pi (n( si )) = It is total production materials

authorize the production kanban card

with total no of work stations

K i production of

(T ) = Start times of materials between two

work stations.

workstation S , initially, in workstation materials part

actual time (t ) and delay time (T ) .

(t )

= End times of materials between two

work stations.

In eq. (1), we have calculate the inventory and

production information from M i . The Pi (n( si )) is part
cell of finish production with kanban card demand
arrives to the system, it is immediately transmitted to all
cells adding on an authorization card (kanban) for the
production of a new part, it is given in eq. (2). The

Algorithm Procedure
Begin

for each object

materials demand timing of the Suppliers t is found

by the difference of the starting time with the starting
workstation of the materials is computed with the help
of the above eq. (3).

for each

find

find

Mi

nS

from set of

Pi nsi

nl

for other

processing (l )

Input:
Dataset

Pi nsi ns t i ns Ti

RM RM 1 , RM 2 ,..., RM N ,

find

Output: Pre-proces sed data

Parameters:

(t i )

st sT
i

end for

M i = Materials ID
find

Mi

= Order ID

Mi

A = Assembly Delay
S = Number of workstation

Pi ns
t T

end for
End

L = Location of workstation

(t i ) = Order timing
(Ti )

select

Pseudo code:

n(O)

RM i

III. RESULTS

= Dispatch timing

The following results were obtained for the

developed Mathematical model and the simulation
results obtained for various parameters with Matlab
Simulink.

nl = Number of workstation location

d j = Starting timing of the assemble timing

ISSN : 2319 3182, Volume-2, Issue-4, 2013

International Journal on Theoretical and Applied Research in Mechanical Engineering (IJTARME)

Fig. 6: Completed Order

Fig. 3: Part A storage materials

IV. CONCLUSION
In this paper, the authors used Simulink and
mathematical model with regression modeling technique
for evaluation of the performance of pull production
control mechanisms, in multi production line, multi
stage assembly manufacturing system with kanban
control system. The authors had drawn the conclusion
that the proposed Simulink model and regression model
have been observed to be performing better to other
control system. Finally, it was noticed that there is no
significance and effect of degree of imbalance in the
system.
V. REFERENCES
Fig. 4: Part B storage materials

Fig. 5: Product demand by system

[1]

Spearman, M.L., Woodruff, D.L., Hopp, W.J.,

CONWIP. A pull alternative to Kanban,
International Journal of Production Research,
Vol. 28, No. 5, pp. 879894, 1999.

[2]

Framinan, J.M., Gonzalez, P.L., Ruiz-Usano, R.,

The CONWIP production control system:
review and research issues, Production Planning
and Control, Vol. 14, No. 3, pp.255265, 2003.

[3]

O. Srikanth, B.V. Ramana Murty and N. Selvaraj

Evaluation of Performance Parameters by
Hybrid Modelling of Pull Production Control
System and by Simulation Approach,
International Journal of Emerging Technology
and Advanced Engineering, Vol. 3, No. 2, 2013.

[4]

Jose M. Framinan, Pedro L. Gonzlez and Rafael

Ruiz-Usano, Dynamic card controlling in a
CONWIP system, International Journal of
Production Economics, Vol. 99, No. 12,
pp.102116, 2006.
ISSN : 2319 3182, Volume-2, Issue-4, 2013

International Journal on Theoretical and Applied Research in Mechanical Engineering (IJTARME)

[5]

Jan-Arne Pettersen and Anders Segerstedt

Restricted work-in-process: A study of
differences between Kanban and CONWIP,
International Journal of Production Economics,
Vol. 118, No. 1, pp. 199207, 2009.

[9]

Arianna Alfieri and Andrea Matta Mathematical

programming formulations for approximate
simulation of multistage production systems,
European Journal of Operational Research, Vol.
219, No.1, pp. 773783, 2012.

[6]

Enns, Silvanus T. and Rogers, Paul, Clarifying

CONWIP versus push system behavior using
simulation, In preceding of Simulation
Conference, Austin, TX, pp. 1867 1872, 2008.

[8]

[7]

Mark L. Spearman and Michael A. Zazanis Push

and Pull Production Systems: Issues and
Comparisons, Operations Research, Vol. 40, No.
3, pp. 521-532, 1992.

P. Lavoie, A. Gharbi and J.-P. Kenne A

comparative study of pull control mechanisms for
unreliable homogenous transfer lines, Int. J.
Production Economics, Vol. 124, No. 5, pp. 241
251, 2010.

[9]

Katsuhisa Ohno, The optimal control of just-intime-based production and distribution systems
and performance comparisons with optimized
pull systems, European Journal of Operational
Research, Vol. 213, No.2, pp. 124133, 2011.

[8]

Bars Selcuk Adaptive lead time quotation in

a pull production system with lead time
responsive demand, Journal of Manufacturing
Systems, Vol. 32, No.4, pp. 138 146, 2013.

[10] Angappa Gunasekaran, kee-hung Laib and T.C.

Edwin Cheng Responsive supply chain: A
competitive strategy in a networked economy,
Omega, Vol. 36, No. 4, pp. 549564, 2008.

ISSN : 2319 3182, Volume-2, Issue-4, 2013